671017.pdf

may lead to computational difficulties during the calculation
of error function derivatives [ 4 – 6 ].
In this research, a new systematic search technique is
proposed on the basis of genetic algorithm (GA) [ 7 – 12 ]and
Ta g u c h i m e t h o d [ 13 – 16 ]. GA as a well-known metaheuristic
algorithm can be utilized to calibrate any soil constitutive
model by means of the results obtained from any laboratory
test and/or in situ experiment [ 2 ]. GA needs only an objective
function rather than its derivatives. In this way, the shortcom-
ings of classical methods will be eliminated as a result. How-
ever, in order to decrease the computational time, sensitivity
analyses are required to select only the dominant parameters
when the input parameters affecting the mechanical behavior
are numerous. In this study, sensitivity analyses are carried
out systematically using the well-known Taguchi method.
This method which is conventionally used for the design of
laboratory experiments can be treated as a modern technique
in geotechnical application.
Genichi Taguchi, who first introduced this method dur-
ing the late 1940s, utilized the conventional statistical tools in
a simplified form by identifying a set of stringent guidelines
for experiment layout and the analysis of results [ 13 ]. He made
an applicable method for design and analysis of factorial
experiments which is mainly used in quality engineering.
Thismethod,wellknownforitsindustrialapplicationsto
identify sensitive parameters for a given target, has fewer
applications in geotechnics, particularly on material property
identification [ 13 ].
In this paper, the results of pressuremeter tests [ 17 , 18 ]
whichareperformedonclayeysandin“LeRheu”site
locatedinFrancehavebeenadoptedforcalibrationofsoil
constitutive model [ 19 ]. The proposed method for inverse
calibration is expressed using a special example which entails
the푃−Δ푉/푉 0 curve obtained by pressuremeter test in a
particular depth [ 20 ]. The constitutive law of this soil is
assumedtobeHSmodelduetothebehaviorthatisexhibited
during laboratory results. Thereafter, the inverse calibration
is repeated with the reduced number of input parameters,
obtained from the Taguchi method.

2. Specifications of the Soil in ‘‘Le Rheu’’ Site

The site is located in the west part of France, in a region
called“LeRheu.”Thesoilofthissitecontainsreddishsand
for tens of meters. Several in situ and laboratory tests have
been performed on this soil to identify its mechanical and
engineering characteristics. The main reason for selection
of this site in current research is the uniformity of the soil
type in different depths and the existence of water table at
very low levels. These conditions reduce the complexity of
modeling process and let all efforts be concentrated on the
mathematical solution for inverse calibration.
The results of pressuremeter tests are available at three
points of B4, P1, and P2 (Figure 1) in various depths of 2 m,
3m,4m,and5m[ 19 ]. However, in this study, only the curve
related to point B4 at the depth of 2 m was selected.Figure 1
illustrates the results of tests at point B4 in the form of푃−
Δ푉/푉 0 curve.

1000

500

0

0

0. 50 1. 00 1. 50

P(

kPa)

Test B

Test P

Test P

V/V 0

Figure 1: Pressuremeter curves at a depth of 2 m after being modi-

fied by lift-off method.

Deviatoric stress

E 50

1

1

Eur

Asymptote

Failure line

Axial strain- 1

qa

qf

Figure 2: Hyperbolic stress-strain relation in primary loading for a

standard drained triaxial test.

3. Hardening Soil Model

The hardening soil model is an advanced model for sim-

ulating the behavior of both soft and stiff soils. When

subjected to primary deviatoric loading, the soil shows a

decreasing stiffness and simultaneously irreversible plastic

strains developing. In the special case of a drained triaxial

test, the observed relationship between the axial strain and

the deviatoric stress can be well approximated by a hyperbola

function, as (Figure 2):

−휀 1 =

1

2퐸 50

푞

1−푞/푞푎

for푞<푞푓. (1)

In contrast to an elastic-perfectly plastic model, the

yield surface of a hardening plasticity model is not fixed

in principal stress space, but it can expand due to plastic

straining. Distinction can be made between two main types of

hardening, namely, shear hardening and compression hard-

ening. Shear hardening is used to model irreversible strains